CN104809467B - A kind of air quality index method of estimation based on dark primary priori - Google Patents
A kind of air quality index method of estimation based on dark primary priori Download PDFInfo
- Publication number
- CN104809467B CN104809467B CN201510181885.4A CN201510181885A CN104809467B CN 104809467 B CN104809467 B CN 104809467B CN 201510181885 A CN201510181885 A CN 201510181885A CN 104809467 B CN104809467 B CN 104809467B
- Authority
- CN
- China
- Prior art keywords
- air quality
- quality index
- characteristic parameter
- sample
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
Abstract
A kind of air quality index method of estimation based on dark primary priori, this method calculates the characteristic parameter of Same Scene hypograph sequence according to dark primary priori theoretical, and the mathematical model of relationship between characteristic parameter and air quality index is established, so as to fulfill the estimation to air quality index；Idiographic flow includes, capturing sample image；Calculate sample image characteristic parameter, down-sampling and the pixel average for calculating N width down-sampled images tri- Color Channels of RGB respectively；Calculate the pixel average of dark primary image；Calculate the characteristic parameter of N width sample images；Make characteristic parameter air quality index scatter plot.Nonlinear fitting, exceptional sample are excluded with being fitted again；Air quality index is estimated.Stronger to the adaptability of different scenes, system structure is simple, and measurement cost is relatively low.For the hardware device of whole system only comprising a digital camera and a personal computer, the peripheral hardware devices without any complexity support that transplantability is good, simple easily to realize.
Description
Technical field
The invention belongs to digital picture/video signal processing fields, are related to a kind of air quality based on dark primary priori
Index method of estimation.
Background technology
The economy of high density population and social activities give off a large amount of particulate matter.These particulate matters are suspended in sky for a long time
In gas, atmospheric visibility can be generated and seriously affected.The content concn of particulate in air is higher, and it is tighter just to represent air pollution
Weight.Air quality index can send out atmosphere pollution early warning to people, enable people to the trip of reasonable arrangement oneself, so as to protect
The health of witnesses.Therefore, the monitoring of air quality index at present has become a highly important job.
Traditional air quality index assay method tends to rely on special instrument, and measurement cost is higher.Beijing's air
The data that quality-monitoring center is announced are shown：The cost of one air quality monitoring substation is about at 900,000 yuan --- and 1,000,000 yuan；
300,000 yuan of the close prices of the instrument of one measurement PM2.5 concentration.Even relatively inexpensive small-sized measuring instrument, price are past
Toward also at ten thousand yuan or more.And the volume and quality of measuring instrument are usually larger, and mobility is poor, not readily portable.In order to drop
The measurement cost of low air quality index can more efficiently instruct people's reasonable arrangement to go on a journey, and find one kind more
Easily new way estimates that air quality index is particularly important.
In recent years, the air quality index indirect measurement method based on image procossing receives the concern of people.This kind of
It connects measuring method and avoids composition of air analytic process complicated in direct measuring method, only by analyzing digital camera same
Collected great amount of images sequence in one scene, it is possible to obtain the valuation of air quality index.Relative to traditional air matter
Volume index measuring instrument, digital camera are cheap, easy to carry.Moreover, monitoring camera widely distributed in city
Machine can become the direct sources of image/video data.However, the air quality index based on image procossing measures skill indirectly
Art is still at an early stage, and the technical solution of this respect is not mature enough.
Patent application No. is CN201310141896.0 discloses a kind of haze monitoring method based on computer vision.
This method obtains characterization object mesopic vision characteristic difference using the color, shape, textural characteristics of close-target object and remote object
Feature mix vector, and estimate air quality index whereby.However, this method need to calculate color saturation respectively it is equal
Value, blue component mean value, SIFT feature are counted out, Canny edge detection results, gray level co-occurrence matrixes, wavelet transformation subband system
Many reference quantities, the computation complexities such as number are higher.Moreover, this method needs to preset close-target object and remote object in the scene.
When object ideal, that distance is clearly demarcated is not present in scene, the validity of this method will be restricted.
Patent application No. is CN201310556138.5 discloses a kind of haze concentration prison based on image analysis technology
Survey method and system.This method mainly obtains monitoring result by simulation system, without presetting close-target object and remote object,
But the physical arrangement of the system is sufficiently complex, and system includes computer system, image capture instrument, rangefinder, includes at least three
The parts such as image simulation system, ventilating system, direction selecting controller and the air-supply pipeline of spray chamber.It measures the realization of function and needs
Complicated peripheral equipment is relied on, measurement cost is still higher, and instrument is not readily portable.
Invention content
For above-mentioned technology there are the problem of, the present invention proposes a kind of air quality index based on dark primary priori and estimates
Meter method.This method calculates the characteristic parameter of Same Scene hypograph sequence according to dark primary priori theoretical, and establishes feature
The mathematical model of relationship between parameter and air quality index, so as to fulfill the estimation to air quality index.
To achieve the above object, the present invention uses following technical scheme：
The flow chart of air quality index method of estimation proposed by the present invention based on dark primary priori is as shown in Figure 1, tool
Body includes the following steps：
Step 1, capturing sample image.
Under the conditions of different air quality indexs, in the sample graph that same Outdoor Scene is acquired in daily fixed time period
Picture, and record the corresponding air quality index Instrument measuring value of each image.N number of sample (N is acquired in total>100), acquisition time
Interval should be greater than 1 hour.Sample image is directed to outdoor fixation monitor camera.
Step 2, sample image characteristic parameter is calculated.
Step 2.1, down-sampling.Down-sampling processing is carried out to whole N width sample images, obtain size be unified for 160 ×
90 N width down-sampled images.
Step 2.2, the pixel average of N width down-sampled images tri- Color Channels of RGB is calculated respectively.The step is by following formula
It realizes：
In formula, i represents that the abscissa of some pixel, j represent the ordinate of some pixel, IR(i, j) represents down-sampling figure
The R channel pixel values of some pixel, I as inG(i, j) represents the G channel pixel values of some pixel in down-sampled images, IB(i,j)
Represent the channel B pixel value of some pixel in down-sampled images, mean represents that the pixel of required tri- Color Channels of RGB is averaged
Value.After calculating, N number of pixel average mean is obtained.
Step 2.3, the dark primary image of N width down-sampled images is obtained respectively.The step is realized by following formula：
In formula, Ic(i, j) represents some pixel in down-sampled images, and c represents tri- Color Channels of RGB,
Expression takes the minimum pixel value of tri- Color Channels of RGB,It represents with the mini-value filtering that Ω (x) is window, in this method
The size of Ω (x) is 5 × 5, JdarkThe dark primary image that (i, j) expression acquires.Dark primary image Jdark(i, j) is a width gray scale
Image.After the step carries out, N width dark primary images can be obtained.
Step 2.4, the pixel average of dark primary image is calculated.The step is realized by following formula：
In formula, Jdark(i, j) represents dark primary image, meandarkRepresent the pixel average of dark primary image.It has been calculated
Bi Hou obtains N number of dark primary image pixel average value meandark。
Step 2.5, the characteristic parameter of N width sample images is calculated.The step is realized by following formula：
In formula, meandarkRepresent the pixel average of dark primary image, mean represents tri- face of RGB of down-sampled images
The pixel average of chrominance channel, γ represent the characteristic parameter of sample image.It can thus be concluded that the characteristic parameter of N number of sample image.
Step 3, characteristic parameter-air quality index scatter plot is made.
Air quality index (AQI) composition N group numbers corresponding the characteristic parameter γ of all sample images is right, flat
Characteristic parameter and the relationship scatter plot of air quality index are made in the rectangular coordinate system of face.
Step 4, nonlinear fitting.
According to minimum mean square error criterion, nonlinear fitting is carried out to scatter plot with single argument power function, is obtained non-linear
The parameters of fit mathematics model.Model of fit is shown below：
In formula, γ represents the characteristic parameter of sample image, and a represents the constant term in model, and b represents the high math power in model
Term coefficient, c represent the fitting power of model,Represent the valuation of air quality index.After step 4 is finished, respectively
To the value of a, b, c.
Step 5, exceptional sample is excluded with being fitted again.
Step 5.1, the air quality index estimated value of sample image is calculated using model of fit.By the feature of sample image
In the mathematical modulo pattern (5) that parameter substitutes into, so as to calculate the corresponding air quality index valuation of each characteristic parameter
Step 5.2, the estimation absolute error of each sample is calculated.The step is realized by following formula：
In formula,For the air quality index valuation that model of fit calculates, AQI is the instrument of air quality index
Measured value, absolute errors of the e between valuation and measured value.
Step 5.3, exceptional sample is excluded with being fitted again.Sample point of the absolute error more than 50 from total sample is concentrated and is arranged
It removes, then performs step 4, obtain new model of fit.
Step 5.4, step 5.1 is repeated to 5.3, until the absolute error of all sample points is respectively less than 50, this
When obtain final characteristic parameter-air quality index relational model.
Step 6, air quality index is estimated.
The scene image is acquired, calculates its characteristic parameter, and characteristic parameter is substituted into characteristic parameter-air quality index and is closed
It is model, you can obtain the valuation of air quality index.It needs again to estimate air quality index under Same Scene
When, only it need to repeat step 6.
Compared with prior art, the present invention has following apparent advantage and beneficial effect.
1. the adaptability of pair different scenes is stronger, independent of close-target object preset in scene and remote object, to family
Outer scene does not have specific limitation, and versatility is stronger；
2. system structure is simple, measurement cost is relatively low.The hardware device of whole system only comprising digital camera and
One personal computer, the peripheral hardware devices without any complexity support that transplantability is good, simple easily to realize；
3. characteristic parameter is determined according to dark primary priori, without calculating the calculation amounts phase such as SIFT feature, wavelet coefficient
To larger parameter, computational efficiency is high, and algorithm speed is fast；
4. the estimation precision of air quality index is high.At present, the measurement accuracy of small air performance figure measuring instrument is about
It is ± 10%.Air quality index is estimated using method proposed by the present invention, average relative error is about 14.53%, accuracy
It is higher, there is highly important reference value to the trip arrangement of people.
Description of the drawings
Fig. 1 is the flow chart of air quality index method of estimation according to the present invention；
Fig. 2 is the absolute error between the instrument measurements and model estimate value that the present invention calculates in fit procedure,
(a) absolute error for fitting for the first time, (b) are the absolute error of second of fitting, and (c) is the exhausted of third time fitting (final)
To error；
Fig. 3 is characteristic parameter-air quality index model of fit that the present invention obtains, and (a) fitting obtains for the first time
Model, (b) are to be fitted obtained model for the second time, and (c), which is that third time is (final), is fitted obtained model.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be further described.
1. capturing sample image
First, under the conditions of different air quality indexs, in daily 8:00 to 17:The same outdoor field of acquisition in 00 period
The sample image of scape, and record the Instrument measuring value of the corresponding air quality index of each image.The sampling time of sample image
Interval is at least 1 hour, can reduce the repeatability of sample in this way.According to the method described above, at least 100 samples are acquired.These
Sample should cover entire air quality index range (about 0 --- 350) as far as possible.
2. calculate characteristic parameter
Collected all sample images are normalized to the image that size is 160 × 90, obtain sample of the same size
Image.Next, calculate the pixel average of tri- Color Channels of RGB of sample image.Then, it is corresponding to acquire sample image
Dark primary image calculates the pixel average of dark primary image.Finally, it is the pixel average of each dark primary image is right with it
The pixel average of tri- Color Channels of RGB of sample is answered to make ratio, obtains the characteristic parameter of all samples.
3. establish the mathematical model of relationship between characteristic parameter-air quality index
It is several right that the corresponding air quality index of the characteristic parameter of sample is formed, and makes characteristic parameter and air quality
The relationship scatter plot of index.According to minimum mean square error criterion, nonlinear fitting is carried out to scatter plot with single argument power function, is obtained
To nonlinear mathematical model.
4. exceptional sample is excluded with being fitted again
The mathematical model that the characteristic parameter of all samples is substituted into calculates the corresponding air quality of each characteristic parameter
Index estimation.Next, the absolute error between computation model valuation and instrument measurements.If the valuation of sample and measured value it
Between absolute error be more than 50, then it is assumed that these samples belong to exceptional sample, and these samples from total sample are concentrated and are excluded.Row
After exceptional sample, nonlinear fitting is carried out according to step 3 again and exceptional sample excludes.The process is constantly repeated, until institute
The absolute error for having sample is respectively less than 50, can obtain final characteristic parameter-air quality index relational model at this time.
5. air quality index is estimated
Single width scene image is acquired, calculates the characteristic parameter of the image, characteristic parameter is then substituted into characteristic parameter-sky
Makings volume index model of fit, you can obtain the valuation of air quality index.
Shown in Fig. 2 is the absolute error that method proposed by the present invention obtains in step 5.2.Scheme (a), (b), (c) point
It is not first time, the absolute error calculated after fitting for the second time and for the third time, it is observed that fitting number is more, absolutely
Error is fewer more than 50 sample size.For the last time after (third time) fitting, absolute error is below 50.
Shown in Fig. 3 is characteristic parameter-air quality index relational model that method proposed by the present invention obtains.Figure (a),
(b), (c) is first time, the model obtained after fitting for the second time and for the third time respectively.The final mask that figure (c) is.
Table 1 illustrates the air quality index data estimated using the present invention.
The AQI estimated results of 1 present invention of table
Claims (1)
1. a kind of air quality index method of estimation based on dark primary priori, it is characterised in that：This method specifically includes following
Step：
Step 1, capturing sample image；
Under the conditions of different air quality indexs, in the sample image that same Outdoor Scene is acquired in daily fixed time period, and
Record the corresponding air quality index Instrument measuring value of each image；N number of sample (N is acquired in total>100), acquisition time interval
It should be greater than 1 hour；Sample image is directed to outdoor fixation monitor camera；
Step 2, sample image characteristic parameter is calculated；
Step 2.1, down-sampling；Down-sampling processing is carried out to whole N width sample images, obtains the N that size is unified for 160 × 90
Width down-sampled images；
Step 2.2, the pixel average of N width down-sampled images tri- Color Channels of RGB is calculated respectively；The step is by following formula reality
It is existing：
In formula, i represents that the abscissa of some pixel, j represent the ordinate of some pixel, IR(i, j) represents certain in down-sampled images
The R channel pixel values of a pixel, IG(i, j) represents the G channel pixel values of some pixel in down-sampled images, IBUnder (i, j) is represented
The channel B pixel value of some pixel in sampled images, mean represent the pixel average of required tri- Color Channels of RGB；Meter
After calculation, N number of pixel average mean is obtained；
Step 2.3, the dark primary image of N width down-sampled images is obtained respectively；The step is realized by following formula：
In formula, Ic(i, j) represents some pixel in down-sampled images, and C represents tri- Color Channels of RGB,Expression takes
The minimum pixel value of tri- Color Channels of RGB,It represents with the mini-value filtering that Ω (x) is window, Ω (x) in this method
Size is 5 × 5, JdarkThe dark primary image that (i, j) expression acquires；Dark primary image Jdark(i, j) is a width gray level image；It should
After step carries out, N width dark primary images can be obtained；
Step 2.4, the pixel average of dark primary image is calculated；The step is realized by following formula：
In formula, Jdark(i, j) represents dark primary image, meandarkRepresent the pixel average of dark primary image；After calculating,
Obtain N number of dark primary image pixel average value meandark；
Step 2.5, the characteristic parameter of N width sample images is calculated；The step is realized by following formula：
In formula, meandarkRepresent the pixel average of dark primary image, mean represents tri- Color Channels of RGB of down-sampled images
Pixel average, γ represent sample image characteristic parameter；It can thus be concluded that the characteristic parameter of N number of sample image；
Step 3, characteristic parameter-air quality index scatter plot is made；
Air quality index (AQI) composition N group numbers corresponding the characteristic parameter γ of all sample images is right, it is straight in plane
Characteristic parameter and the relationship scatter plot of air quality index are made in angular coordinate system；
Step 4, nonlinear fitting；
According to minimum mean square error criterion, nonlinear fitting is carried out to scatter plot with single argument power function, obtains nonlinear fitting
The parameters of mathematical model；Model of fit is shown below：
In formula, γ represents the characteristic parameter of sample image, and a represents the constant term in model, and b represents the high math power term system in model
Number, c represent the fitting power of model,Represent the valuation of air quality index；After step 4 is finished, respectively obtain a,
B, the value of c；
Step 5, exceptional sample is excluded with being fitted again；
Step 5.1, the air quality index estimated value of sample image is calculated using model of fit；By the characteristic parameter of sample image
It substitutes into obtained mathematical modulo pattern (5), so as to calculate the corresponding air quality index valuation of each characteristic parameter
Step 5.2, the estimation absolute error of each sample is calculated；The step is realized by following formula：
In formula,For the air quality index valuation that model of fit calculates, AQI is the instrument measurements of air quality index,
Absolute errors of the e between valuation and measured value；
Step 5.3, exceptional sample is excluded with being fitted again；Sample point of the absolute error more than 50 from total sample is concentrated and is excluded, so
Step 4 is performed afterwards, obtains new model of fit；
Step 5.4, step 5.1 is repeated to 5.3, until the absolute error of all sample points is respectively less than 50, at this time
To final characteristic parameter-air quality index relational model；
Step 6, air quality index is estimated；
The scene image is acquired, calculates its characteristic parameter, and characteristic parameter is substituted into characteristic parameter-air quality index relationship mould
Type, you can obtain the valuation of air quality index；When needing again to estimate air quality index under Same Scene, only
It need to repeat step 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510181885.4A CN104809467B (en) | 2015-04-16 | 2015-04-16 | A kind of air quality index method of estimation based on dark primary priori |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510181885.4A CN104809467B (en) | 2015-04-16 | 2015-04-16 | A kind of air quality index method of estimation based on dark primary priori |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104809467A CN104809467A (en) | 2015-07-29 |
| CN104809467B true CN104809467B (en) | 2018-07-06 |
Family
ID=53694278
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510181885.4A Expired - Fee Related CN104809467B (en) | 2015-04-16 | 2015-04-16 | A kind of air quality index method of estimation based on dark primary priori |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809467B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020088A (en) * | 2014-05-15 | 2014-09-03 | 清华大学深圳研究生院 | Method for obtaining concentration of particles in air based on image processing |
| CN104217404A (en) * | 2014-08-27 | 2014-12-17 | 华南农业大学 | Video image sharpness processing method in fog and haze day and device thereof |
| CN104517126A (en) * | 2014-12-26 | 2015-04-15 | 北京邮电大学 | Air quality assessment method based on image analysis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2005101312A2 (en) * | 2004-04-13 | 2005-10-27 | Aic | Fractal skr-method for evaluating image quality |
-
2015
- 2015-04-16 CN CN201510181885.4A patent/CN104809467B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104020088A (en) * | 2014-05-15 | 2014-09-03 | 清华大学深圳研究生院 | Method for obtaining concentration of particles in air based on image processing |
| CN104217404A (en) * | 2014-08-27 | 2014-12-17 | 华南农业大学 | Video image sharpness processing method in fog and haze day and device thereof |
| CN104517126A (en) * | 2014-12-26 | 2015-04-15 | 北京邮电大学 | Air quality assessment method based on image analysis |
Non-Patent Citations (1)
| Title |
|---|
| 基于改进的暗原色先验理论的监控视频去雾算法;王金艳;《中国优秀硕士学位论文全文数据库·信息科技辑》;20141115(第11期);I138-471 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104809467A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8988317B1 (en) | Depth determination for light field images | |
| RU2012134347A (en) | METHOD AND SYSTEM FOR DETERMINING COLOR BY IMAGE | |
| CN110487737B (en) | Image information extraction and calculation method and system for spectrum detection of smart phone | |
| KR101116462B1 (en) | Method and apparatus for water quality monitoring using remote sensing technique | |
| CN103940714A (en) | Imitated artificial haze monitoring system and method | |
| CN105241826A (en) | Intelligent mobile terminal and food detection method using the same | |
| CN109459136B (en) | Method and device for measuring chromaticity | |
| CN102421009A (en) | Automatic video signal testing method | |
| CN107238442B (en) | Without the detector thermometric method for regulating temperature of temperature stabilizer | |
| CN103581634A (en) | Processing method for image wide dynamic range | |
| CN110044485B (en) | Image type fabric color measuring method | |
| US9188523B2 (en) | System for estimating size distribution and concentration of aerosols in atmospheric region | |
| CN104809467B (en) | A kind of air quality index method of estimation based on dark primary priori | |
| CN108225318B (en) | Image quality-based aviation remote sensing path planning method and system | |
| CN111062926B (en) | Video data processing method, device and storage medium | |
| KR101992403B1 (en) | Skin moisture estimating method using image | |
| CN105704382B (en) | A kind of multichannel CCD camera method for correcting image and device | |
| CN112488997A (en) | Method for detecting and evaluating color reproduction of ancient painting printed matter based on characteristic interpolation | |
| CN106530248A (en) | Method for intelligently detecting scene video noise of transformer station | |
| CN114397094B (en) | Light source color rendering evaluation method based on spectrum reconstruction and color appearance model | |
| CN115937093A (en) | Smoke concentration detection method integrating HSL space and improved dark channel technology | |
| WO2022242608A1 (en) | Object preference memory color obtaining method and preference memory color standard color card | |
| CN113758683B (en) | Camera system post-irradiation spectrum degradation evaluation method based on average color saturation | |
| CN111091601B (en) | PM2.5 index estimation method for real-time daytime outdoor mobile phone image | |
| CN113792616A (en) | Remote meter reading system based on edge calculation and working method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180706 |